As we enter the age of artificial intelligence and automation, how do we ensure autonomous vehicles are safe? What about automated air traffic control, or nuclear power plants fully controlled by computers? This is where control software systems come in. Acting like the conductor of an orchestra, they coordinate the actions of the different parts of an autonomous system based on feedback they receive from those parts. Understanding these interacting feedback loops is vital for developing control systems that we truly trust.
“Autonomous systems are complex for many reasons, but the complexity that I'm looking at results from feedback loops that involve both physics and software,” says Anne-Kathrin Schmuck, who leads the Control Software Systems Group at the Max Planck Institute for Software Systems (MPI SWS) in Kaiserslautern, Germany. She points to the example of air traffic control at an airport, which ensures that planes do not collide during take-off and landing. Fully automated air traffic control is challenging due to the complex interplay of the plane’s flight dynamics, its space and time constraints to avoid collisions, human pilot interventions, and external factors such as changing weather conditions or flight schedules.
“If multiple complex dynamical systems are interacting with each other, you need to make sure that they always react correctly – especially when the lives of humans are at risk. With the ever-increasing autonomy levels and the tighter integration of physics and software in such systems, ensuring scrupulous reliability of all involved feedback mechanisms is becoming almost impossible,” she says.
While each modern aeroplane is equipped with countless feedback-controllers fully automatically regulating their flight dynamics, the problem of assigning landing and starting spots along with non-intersecting flight corridors to each plane is, as of today, still mostly done manually by “an army of humans” sitting in the air traffic control towers ensuring nothing goes wrong. “Because we do not fully understand all involved feedback mechanisms yet, we simply can’t trust fully automated control software enough to hand this task over to a computer, even though it's a seemingly simple task.” Her Control Software Systems Group is therefore doing the fundamental research on feedback mechanisms that could one day underpin trusted autonomous safety critical systems.
Before leading the group, Anne-Kathrin joined the MPI SWS as a postdoctoral researcher, following time studying Engineering Cybernetics in Germany, Canada and Sweden. In 2020 she received a prestigious Emmy Noether Grant to establish her own research group at MPI SWS - an award that has given her the academic freedom to pursue “more risky ideas” that she would otherwise struggle to get third-party funding for. She was well-supported in her application for the grant, she says, because “many faculty members happily shared their experiences with prestigious grant proposals”. Their feedback was extremely useful in shaping her application. “For me, this was really a key step in my career. It forced me to spell out my research vision very concretely, which now allows me to attract strong PhD candidates with well-defined, cutting-edge research problems.”
Researchers from students and interns up to the faculty level are attracted to the Institute because of its reputation, she adds. It is the perfect place for her research, as Anne-Kathrin can combine her expertise in control engineering with the software-centred expertise of her colleagues. And while the Institute's smaller size creates a bit of a family feel, its connection to the other 85 Max Planck Institutes - as well as neighbouring research facilities - fosters a wide array of collaborations. Anne-Kathrin’s group is based on a campus in Kaiserslautern, a picturesque city on the edge of the Palatinate Forest. Known as a bit of an international melting pot, English is widely spoken there due to a nearby US air force base. The MPI SWS itself is split across two campuses, with the second an easy train away in Saarbrücken, another small charming city that hugs the French border.
The work isn’t only done on these two campuses though - Anne-Kathrin also gets to make the most of an extensive travel budget available to the Institute's researchers. “This year I travelled to the US for two weeks to see collaborators, in Europe I've been to France, to Belgium, to the Netherlands, to Sweden,” she explains. While travelling for research visits and conferences opens up valuable learning, networking, and collaboration opportunities, she is also grateful for the support she receives as a working mother. The Institute has spots reserved in the German childcare system, which she says can be competitive and resource-constrained. “That of course is amazing in terms of not worrying about this,” she says. Another benefit she highlights is that researchers can fully focus on their research, as there is no teaching requirement at the Institute. “We’re really a base research institute, we can teach if we want to,” she explains. She does teach one class at a neighbouring university, but not being required to teach provides the freedom and flexibility her team needs to focus on the fundamental research that will hopefully make control systems reliable and robust enough to orchestrate our complex, automated world.
The Max Planck Institute for Software Systems is chartered to conduct world-class basic research in all areas related to the design, analysis, modeling, implementation and evaluation of complex software systems.See all current vacancies
The Max Planck Institute for Software Systems is chartered to conduct world-class basic research in all areas related to the design, analysis, modeling, implementation and evaluation of complex software systems.Besøg arbejdsgiversiden
Dr. Schmuck is a faculty member at the Max Planck Institute for Software Systems (MPI SWS) in Kaiserslautern, Germany. She received a prestigious Emmy Noether Grant and is now leading the Control Software Systems Group at MPI SWS.